Abstract
This thesis examines the difference in value and risk relevance of historical cost and fair value accounting information pertaining to banks in the United States of America (U.S.) between 1986 and 2013. Under U.S. Generally Accepted Accounting Principles (GAAP), the difference between historical cost and fair value information is reported as unrealised gains and losses for financial assets categorised as available-for-sale and subsumed under other comprehensive income.
Edward and Bell’s (1961) reporting framework is used as the conceptual basis to reconcile historical costs with fair values and to explain the distinction between net income and other comprehensive income based on operating gains versus holding gains. Through a combination of cross-sectional and time-series regressions, the analysis of the value relevance uses Falta and Willett’s (2013) multiplicative market-accounting model instead of an additive linear model with the risk relevance analysis based on the work of Lubberink and Willett (2017).
Findings indicate that the statistical difference between historical cost and fair value for available-for-sale securities and securities held-to-maturity is not significant. The results of the cross-sectional analysis of the long-run market-book value relationship suggests that the book value of net assets and net income are relevant for market returns in the long run. Fair value information contained in other comprehensive income calculations is not relevant except for a time during the global financial crisis.
The results of single variable error correction modelling indicate that, in the short-run, markets partially (three to four years) react to changes in the book value of U.S. banks, though not in the case of every bank. The results of a multivariable error correction model indicate that the market over-reacts to changes in book value in the bank sample and that the market pays greater attention to net asset values in the banking industry than it does in other industries. Similar to the results in the cross-sectional analysis, elasticities on book value and net income are highly and negatively correlated. Other comprehensive income has either a small, or a negligible, elasticity in most years. Together the estimates show a strong long-term mean reversion in returns and suggest an adjustment period of somewhere between three and four years of market value to underlying fundamental accounting information.
Under the assumption that the book-to-market ratio is a direct measure of risk the findings indicate that markets consider net income information to be more risk relevant than other comprehensive income information. The two main contributors to the behaviour of the ‘long-run’ risk premium are net income and dividends and tend to act in opposition to one another. Book value growth and other comprehensive income contributed little to banks’ long-run risk premium’ during the period studied. This suggests that the fair values included in the assessment of other comprehensive income are not treated by the market as being informative about risk any more than they are informative about market returns. In other words, the findings suggest that fair values have little incremental information about the performance or value relevance of the banks.